Pattern Mining of Aero-Engine Performance Degradation Based on EMD and KNN-FSFDP

The efficiency of current aero-engine remaining life prediction methods has room for improvement. This paper focuses on analysis of aero-engine exhaust gas temperature margin (EGTM) time-series data and introduces a pattern mining method of aero-engine performance degradation based on Empirical Mode Decomposition (EMD) and k-nearest neighbors optimized clustering by fast search and find of density peaks (KNN-FSFDP). First, we remove the gross error of EGTM time-series data by using box-plot. Then EGTM time-series data are divided into a finite number of Intrinsic Mode Function (IMF) by using EMD. IMF can be divided into noise signals and useful signals because its energy density multiplied by its average period is a constant. We filter the noise signals by Savitzky-Golay filter and recombined the useful signals and the filtered noise signals. After EGTM time-series data are clustered by KNN-FSFDP, we determine the pattern of aero-engine performance degradation. Then aero-engine remaining life can be predicted by the pattern. At the end, the method in this paper is validated by the actual data of aero-engine. The result shows that the method in this paper has better efficiency, so the method can provide support for aero-engine removal time prediction and maintenance plan optimization.